15:01:55 <haleyb> #startmeeting neutron_dvr
15:01:56 <openstack> Meeting started Wed Mar  1 15:01:55 2017 UTC and is due to finish in 60 minutes.  The chair is haleyb. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:58 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:02:00 <openstack> The meeting name has been set to 'neutron_dvr'
15:02:05 <haleyb> #chair Swami
15:02:06 <openstack> Current chairs: Swami haleyb
15:02:18 <haleyb> #topic Announcements
15:03:03 <haleyb> Swami: did you have any announcements?  Any DVR things that were discussed at the PTG?
15:03:33 <Swami> haleyb: hi welcome back
15:03:49 <Swami> yes just wanted to go over the PTG discussions that we had on DVR
15:04:04 <haleyb> thanks, the floor is yours
15:04:13 <Swami> PTG DVR discussions
15:04:28 <Swami> We had four topics that we discussed regarding the DVR.
15:04:46 <Swami> 1. Handling allowed_address_pairs with DVR.
15:05:34 <Swami> 2. Adding an config option to allow the Floatingips to be configured in the network node. This is to force all the North-South traffic to go through the centralized node.
15:06:26 <Swami> 2.(a). Proactively check the agent health before you schedule a floatingip to a compute node, and if the agent is not running (dead) then move the floatingip to the centralized node.
15:08:08 <Swami> 3. Distributed SNAT. We discussed the issues about the inability to share the SNAT namespace between the tenants and we all agreed that if the customer is willing to burn the public ip addresses, then SNAT can be distributed. Still needs investigation if we need a config option or can we run 'dvr_snat' agent on all nodes.
15:09:52 <Swami> 4. There was another suggestion about implementing the floatingip and SNAT using the openflow rules and getting rid of the namespaces that we use today. This one was suggested by 'ajo'. He trying out a prototype and once it is working, he might push in a spec or an RFE.
15:10:23 <Swami> #link https://bugs.launchpad.net/neutron/+bug/1583694 - Allowed address pair RFE
15:10:23 <openstack> Launchpad bug 1583694 in neutron "[RFE] DVR support for Allowed_address_pair port that are bound to multiple ACTIVE VM ports" [Wishlist,In progress] - Assigned to Swaminathan Vasudevan (swaminathan-vasudevan)
15:11:09 <Swami> #link https://bugs.launchpad.net/neutron/+bug/1667877 - RFE for the config option on Floatingip to be centralized or circumstantial.
15:11:09 <openstack> Launchpad bug 1667877 in neutron "[RFE] DVR support for Configuring Floatingips in Network Node or in the Compute Node based on Config option." [Wishlist,Confirmed] - Assigned to Miguel Lavalle (minsel)
15:11:18 <haleyb> I could see #4 working now that OVS has conntrack, since that was probably one of the reasons we never looked at it before
15:11:32 <Swami> haleyb: yes.
15:12:12 <haleyb> I am curious about #2 though
15:12:19 <Swami> ajo will be working on item '4' but he mentioned that he might need some help from ous.
15:13:16 <haleyb> why is having floating IP on the NN a good thing?  that's typically the case you want to go direct
15:13:22 <Swami> haleyb: #2 megulle might help us in item #2 to come up with an function that can check the agent state or status.
15:14:13 <Swami> haleyb: kevin mentioned that there are some customers who are looking for using the centralized node for Floatingip and that is the only reason.
15:14:24 <haleyb> if agent is broken i can see it, but if things are in a half-configured state and it dies, will that even work?
15:14:50 <Swami> haleyb: we need to investigate.
15:14:51 <haleyb> Swami: they can always disable DVR
15:15:13 <Swami> haleyb: by disabling DVR they might not get East-West only.
15:15:39 <Swami> haleyb: The reason may be if someone wants just East-West and not North-South this would help.
15:16:20 <Swami> haleyb: I agree with your point if all the issues related to IPv6 is solved for the compute node FIP namespace, then they don't need it to be centralized.
15:16:48 <haleyb> that is going to severly limit bandwidth as we know too well, I guess I always like to see the reasoning behind the "why" there
15:17:40 <Swami> haleyb: Yes that was also discussed and it was said that it is upto the customers to choose which way they wanted and providing an option would be nice.
15:19:31 <haleyb> i'm curious if that's "deployers" or "customers" - deployers should be able to, customers shouldn't need to imo
15:20:27 <Swami> haleyb: yes I meant the deployers not the tenants.
15:20:39 <haleyb> ok, that makes me feel better :)
15:22:27 <haleyb> Swami: regarding address pairs, what was the decision?  or maybe it's in the bug
15:22:39 <Swami> here is the summary that kevin sent out that captures all the four items.
15:22:47 <Swami> #link http://lists.openstack.org/pipermail/openstack-dev/2017-February/113032.html
15:23:27 <Swami> haleyb: for address pairs kevin is ok with the current proposal so I have resubmitted those two patches.
15:23:37 <haleyb> Ah, ok, I have not read all email this morning
15:24:05 <Swami> #link https://review.openstack.org/#/c/437970/
15:24:19 <Swami> #link https://review.openstack.org/#/c/437986/
15:24:46 <Swami> I also had discussion with the Octavia folks on these patches and requested them to test it with their use cases.
15:26:16 <Swami> haleyb: that's all I had on the PTG update with respect to DVR.
15:26:27 <Swami> haleyb: You missed our Neutron Team Photo
15:26:55 <haleyb> maybe they can photoshop me in :)
15:27:05 <Swami> #link https://souminathan.smugmug.com/OpenStack-Neutron-PTG-Atlanta-2017/i-HBdQdZ7/A
15:27:18 <Swami> haleyb: will do
15:28:30 <Swami> That's all I had on the PTG update.
15:28:45 <haleyb> Well, it sounds like you were productive, sorry i had to miss it
15:28:57 <haleyb> but the beach was nice too :)
15:29:03 <Swami> haleyb: no problem
15:29:16 <Swami> haleyb: May be you should send some good pictures.
15:29:54 <Swami> haleyb: Was the weather good in Mexico.
15:30:27 <haleyb> I will get some together, weather was 85 every day, almost perfect weather, only rained a few hours
15:30:39 <Swami> haleyb: that is great
15:31:02 <haleyb> work stinks, i need to retire :)
15:31:39 <haleyb> moving on though...
15:31:43 <haleyb> #topic Bugs
15:32:01 <haleyb> I just see the RFE is new
15:32:17 <Swami_> Sorry got disconnected again
15:32:21 <Swami_> haleyb: which RFE you are talking about
15:32:35 <haleyb> https://bugs.launchpad.net/neutron/+bug/1667877
15:32:35 <openstack> Launchpad bug 1667877 in neutron "[RFE] DVR support for Configuring Floatingips in Network Node or in the Compute Node based on Config option." [Wishlist,Confirmed] - Assigned to Miguel Lavalle (minsel)
15:32:52 <Swami_> haleyb: yes I created it after the PTF meeting.
15:33:01 <Swami_> s/PTF/PTG
15:34:47 <haleyb> any other bugs to discuss?  i don't know what needs reviewing just by looking
15:35:49 <Swami_> haleyb: Yes there are no new bugs this week.
15:36:11 <Swami_> haleyb: We might have to push the Fast path exit patches.
15:36:32 <Swami_> #link https://review.openstack.org/#/c/283757/
15:36:46 <Swami_> #link https://review.openstack.org/#/c/355062/
15:37:08 <Swami_> haleyb: I forgot to mention about the DVR multinode job in gate.
15:37:58 <Swami_> haleyb: In PTG we had a discussion and they were planning to get rid of the DVR multinode job and instead have the DVR+HA job in place which is basically multinode with 3 nodes. (2 ctlr and 1 compute)
15:38:00 <haleyb> i need to update the change, looks like failure rate is below 10% right now
15:38:36 <haleyb> oh, i don't think that change merged yet
15:38:53 <Swami_> haleyb: Yes a decision was made to focus and try to merge those patches.
15:39:59 <haleyb> i guess i need to go to more of these meetings
15:40:15 <Swami_> haleyb: It is becoming more productive
15:41:25 <haleyb> I will try and find the DVR+HA reviews, they're on my review page
15:41:32 <Swami_> haleyb: thanks
15:41:57 <Swami_> that's all I have for today's discussion.
15:42:45 <haleyb> I guess we can skip talking about gate as it seems ok, and go right to open discussion
15:42:52 <haleyb> #topic Open Discussion
15:43:12 <haleyb> any other topics to discuss?
15:43:30 <Swami_> haleyb: I heard that john schwarz will not be working on openstack since his directions have changed.
15:44:00 <haleyb> yes, i heard that as well, do not think he is at red hat any more
15:44:03 <Swami_> haleyb: I am not sure if you knew this before, but I had a chat with Anil venkata at the PTG summit.
15:44:12 <Swami_> haleyb: yes that's what I heard.
15:45:00 <haleyb> So is anil the main HA contact now?
15:45:29 <Swami_> haleyb: yes
15:46:38 <haleyb> i do not see him here at the moment
15:47:30 <Swami_> haleyb: yes I don't see him.
15:48:13 <haleyb> i don't have anything else, so going once...
15:49:06 <haleyb> #endmeeting